In-Context Adaptation to Concept Drift for Learned Database Operations

2025-05-07

AI Summary

To address performance degradation of learned operators—such as cardinality estimation—under concept drift in dynamic databases, this paper proposes FLAIR, an online adaptive framework. FLAIR introduces the novel “in-context adaptation” paradigm: instead of parameter updates, it instantaneously constructs a dynamic context (C_t) from query execution feedback and models the prediction function as (f(x mid C_t) o y). The framework integrates a task characterization module, a Bayesian meta-trained dynamic decision engine, and a dynamic context memory mechanism. Experiments on cardinality estimation demonstrate that FLAIR achieves 5.2× faster adaptation and reduces average error by 22.5% compared to state-of-the-art methods. Crucially, it delivers real-time, accurate prediction alignment without any retraining overhead—enabling zero-cost adaptation to evolving data distributions.

Technology Category

• Machine Learning: Online Learning & Bandits
• Data Mining & Knowledge Management: Intelligent Query Processing
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Machine learning has demonstrated transformative potential for database operations, such as query optimization and in-database data analytics. However, dynamic database environments, characterized by frequent updates and evolving data distributions, introduce concept drift, which leads to performance degradation for learned models and limits their practical applicability. Addressing this challenge requires efficient frameworks capable of adapting to shifting concepts while minimizing the overhead of retraining or fine-tuning. In this paper, we propose FLAIR, an online adaptation framework that introduces a new paradigm called extit{in-context adaptation} for learned database operations. FLAIR leverages the inherent property of data systems, i.e., immediate availability of execution results for predictions, to enable dynamic context construction. By formalizing adaptation as $f:(mathbf{x} ,| ,mathcal{C}_t) o mathbf{y}$, with $mathcal{C}_t$ representing a dynamic context memory, FLAIR delivers predictions aligned with the current concept, eliminating the need for runtime parameter optimization. To achieve this, FLAIR integrates two key modules: a Task Featurization Module for encoding task-specific features into standardized representations, and a Dynamic Decision Engine, pre-trained via Bayesian meta-training, to adapt seamlessly using contextual information at runtime. Extensive experiments across key database tasks demonstrate that FLAIR outperforms state-of-the-art baselines, achieving up to 5.2x faster adaptation and reducing error by 22.5% for cardinality estimation.

Problem

• Addressing concept drift in learned database operations
• Minimizing retraining overhead for dynamic database environments
• Enhancing performance of learned models with in-context adaptation

Innovation

• Online adaptation framework for database operations
• Dynamic context memory eliminates runtime parameter optimization
• Pre-trained Bayesian meta-training for seamless adaptation